Motor neurons show significant regenerative capacity

were HB9:GFP⁺ (n=3). This indicates that large HB9:GFP⁺ cells are innervating muscle tissue and therefore are mature motor neurons.

The response of small (< 12 µm diameter) and large (> 12 µm diameter)

HB9:GFP⁺ motor neurons to a lesion was determined (Fig. 5) in an area of 750 µm rostral and caudal to the lesion site. The number of large GFP⁺ cells was significantly reduced at 1 wpl (p = 0.0035, n = 4 vs. 3 animals) and 2 wpl (p=

0.0003, n = 4 vs. 11 animals). After 6 to 8 weeks the number of large motor neurons was increased again to levels that were not significantly different from the unlesioned situation (p = 0.0867, n = 4 unlesioned vs. 6 lesioned animals).

This showed that the original number of mature HB9⁺ motor neurons is

decreased in response to the lesion event. Furthermore it indicates a trend in recovery of the number of large cells.

Numbers of small HB9:GFP⁺ motor neurons responded inversely to the

transection of the spinal cord. A significant increase after 2 wpl (p < 0.0001, n = 4 unlesioned vs. 11 lesioned animals) was followed by a significant decrease in number of small neurons at 6 to 8 weeks (p = 0.0002, n = 11 animals at 2 wpl vs. 6 animals at 6 to 8 wpl).

Fig 5: Dynamic changes in the numbers of HB9:GFP⁺ motor neurons after a lesion. A spinal cord lesion induces an increase in the number of small and a decrease in the number of large motor neurons at 2 wpl. At 6 to 8 wpl, the population of large motor neurons partly recovers, while numbers of small HB9:GFP⁺ cells return to original levels. Stereological counts of HB9:GFP⁺ cells calculated to 1500µm around the lesion site are given.

This transient, more than 43-fold increase, in the number of small HB9:GFP⁺ motor neurons indicates a highly dynamic response in the number of spinal motor neurons to the lesion event. In addition, the time course matches that of the functional recovery, indicating a possible link between motor neuron

regeneration and functional recovery.

Using the islet-1/2 antibody the spatial distribution of differentiating motor neurons was analysed in 14 µm cryosections at increasing distances from the lesion site (Fig. 6). Close to the lesion site (0-250 µm) the number of islet-1/2 positive cell profile counts is highest and significantly increased 2 wpl (p = 0.0253, n = 5 animals each group) compared to unlesioned controls. This corresponds to proliferative activity in the ventricular zone, which is also highest close to the lesion site (Fig. 4 D).

Fig 6: Islet-1/-2 immunohistochemistry confirms an increase in the number of differentiating motor neurons. A: Few large nuclei (arrowhead) are visible in the unlesioned spinal cord. In the lesioned situation, clusters of small islet-1/-2

immunopositive cell nuclei appear in the ventro-lateral spinal cord (arrow). B: Numbers of islet-1/-2 immunopositive cell profiles were determined in cryosections (14 µm in thickness) for the regions indicated, showing a significant increase in islet-1/-2 immunopositive cell profiles around the lesion site. (n = 5 unlesioned animals; n = 5 animals at 2 wpl; p = 0.0253). Bar = 50 µm.

3.1.3.2 Small motor neurons are newly generated after a lesion

To directly address whether motor neurons were newly generated, BrdU was injected into HB9:GFP and islet-1:GFP transgenic animals at 0, 2, and 4 dpl post-lesion and the number of double labelled neurons was determined at 2 wpl.

At 2 wpl there was an increase in the number of small islet-1:GFP positive cells, which was statistically significant compared with the unlesioned situation

(unlesioned: 27 ± 3.9 cells, n = 5 animals, 2 wpl: 870 ± 244.9 cells, n = 4 animals, p = 0.0139). In BrdU injected animals, 184 ± 49.3 small cells (n = 3 animals, p = 0.0104) were double labelled with the transgene and BrdU

immunohistochemistry. In the unlesioned controls no double-labelled cells ( n = 5 animals) were found (Fig. 7).

Fig 7: Newly generated small islet-1:GFP⁺ cells in the lesioned spinal cord. Cross-sections through the spinal cord of unlesioned A: and lesioned B-E: animals at 2 weeks post-lesion are shown. In unlesioned animals only large GFP⁺ cells are

detectable, whereas many smaller GFP⁺ cells are present in the ventrolateral aspect of the lesioned spinal cord. Many of these cells are also BrdU⁺, as indicated by arrows in the higher magnification C-E: of the area boxed in B. Dots outline the ventricle. Bars = 25 μm.

The HB9:GFP transgenic fish confirms these observations: at 2 wpl the small HB9:GFP⁺ cells were increased from 20.0 ± 7.66 in the unlesioned situation (n = 4 animals) to 869.5 ± 106.78 (n = 11 animals, p < 0.0001). In this transgenic fish, 200.0 ± 46.2 cells (n = 7 animals, p = 0.0076) were double-labelled by the transgene and BrdU at 2 wpl (Fig. 8). In the unlesioned spinal cord only one double-labelled motor neuron was observed (n = 4 animals). Even a BrdU injection protocol extended to the maximum number of injections tolerated by the fish (injections at 0, 2, 4, 6, 8 days lesion, analysis at 14 days

post-lesion) did not yield any HB9:GFP⁺/BrdU⁺ cells in unlesioned fish (n = 5

animals). Letting the fish swim in BrdU-treated water in order to label all newly generated cells over the entire time of the experiment does not show sufficient labeling of dividing cells (Dr. Thomas Becker, personal communication). Hence the unlesioned mature spinal cord appears virtually quiescent with respect to motor neuron generation. However, low rates of motor neuron formation may have been missed due to the limited metabolic availability of BrdU.

Fig. 8: Generation of new motor neurons in the lesioned spinal cord. HB9:GFP/BrdU double-labelled neurons are present in the lesioned, but not the unlesioned, ventro-lateral spinal cord. These cells (boxed in upper right and shown in higher magnification in bottom row) bear elaborate processes (arrows) or show ventricular contact

(arrowhead). Bars = 25 µm.

3.1.3.3 Lesion induces cell death

The number of large HB9⁺ motor neurons decreases significantly after a lesion.

We performed TUNEL staining in HB9:GFP transgenic fish at 3 dpl and found TUNEL⁺/HB9:GFP⁺ cells (Fig. 9). The apoptosis marker TUNEL labels the nuclei of cells undergoing programmed cell death (Hewitson et al., 2006).

Fig. 9: Lesion induced apoptosis at 3 dpl. HB9:GFP (green), DAPI nuclear staining (blue) and TUNEL staining (red). Triple labelling indicates apoptotic motor neurons (arrow). Bars: left 15µm, right 8 µm.

3.1.3.4 Different subpopulations of newly generated motor neurons may be present in the lesioned spinal cord

The islet-1:GFP and the HB9:GFP transgenic animals show a similar

distribution of small motor neurons in the ventral horn of the lesioned spinal cord. For islet-1, the transgene expression confirms the expression of the endogenous gene because 89.5 % of the islet-1:GFP⁺ cells were islet-1/2 immunopositive at 2 wpl. The small proportion of cells only labelled by GFP in islet-1:GFP animals may result from higher stability of the GFP than

endogenous islet-1 detected by the antibody. In contrast, a substantial

proportion, 51.7 %, of HB9:GFP⁺ cells were not double-labelled by the islet-1/2 antibody and many cells were exclusively labelled by the islet-1/-2 antibody in both transgenic lines (55.7 % in the HB9:GFP and 35.4 % in the islet-1:GFP fish) (Fig. 10). This suggests heterogeneity among newly generated motor neurons with respect to marker expression (William et al., 2003).

Fig. 10: Partial overlap of islet-1/-2 immunohistochemistry and transgenic motor

neuron markers in the lesioned spinal cord. A: Islet-1:GFP⁺ cells are double-labelled by the islet-1/-2 antibody, confirming specificity of transgene expression. A substantial proportion of HB9:GFP⁺ cells are not double-labelled by the antibody and many cells are only labelled by the islet-1/-2 antibody in both transgenic lines, suggesting that different types of cells were generated after a lesion. Arrows indicate double-labelled neurons, arrowheads indicate neurons only labelled by the transgene and open arrowheads point to cells only labelled by the antibody. B: Summations of all cells counted in six sections (50 μm thickness) per animal from the region of 1.5 mm surrounding the lesion site (n = 3 animals for each transgene) are indicated. Bar = 25 μm.

3.1.3.5 Newly generated motor neurons show terminal differentiation and may be integrated into the spinal network

To determine whether newly generated motor neurons fully matured, expression of ChAT, a marker for terminally differentiated motor neurons (Arvidsson et al., 1997), and coverage of motor neurons by SV2⁺ contacts, a marker for synaptic coverage, was analysed. In the unlesioned situation 80.6%

(n = 3 animals) of the large HB9:GFP⁺ cells expressed ChAT, indicating that the majority of HB9:GFP⁺ cells were mature motor neurons. Small HB9:GFP⁺ cells were rarely found. Furthermore, all ChAT⁺ cells were covered with synapses in the unlesioned spinal cord.

At 2 wpl, small HB9:GFP⁺ neurons were rarely ChAT positive (2.8%, n = 3 animals) and did not receive SV2⁺ contacts (Fig. 11, upper row). Of the large HB9:GFP⁺ cells, 36.4% (n = 3 animals) were double labelled with ChAT and often not covered with SV2 labelled synapses (Fig. 11, middle row). This indicates that most small and some of the large HB9:GFP⁺ neurons were immature at 2 wpl.

To determine whether newly generated motor neurons show terminal differentiation and network integration at later stages of regeneration, BrdU injections at day 0, 2 and 4 were combined with anti-ChAT and anti-SV2 immunohistochemistry. At 6 wpl 29.3 ± 23.14 ChAT⁺ cells/1500µm (n = 3

animals) were also BrdU⁺ and extensively covered with SV2 labelled synapses.

The inset in the lower row indicates that similar cells are part of the typical cytoarchitecture of the unlesioned spinal cord (Fig.11, lower row). These observations are consistent with the hyposthesis that newly generated motor neurons can fully mature and integrate into the spinal network.

Fig. 11: Maturation of newly generated motor neurons. Confocal images of spinal cross-sections are shown (dorsal is up). Clusters of newly generated HB9:GFP⁺ motor neurons are ChAT^- (arrow in top row indicates a ChAT⁺/HB9:GFP^- differentiated motor neuron). Somata (arrow in middle row) and proximal dendrites (arrowheads in middle row) receive few SV2⁺ contacts at 2 wpl. At 6 wpl, ChAT⁺/BrdU⁺ somata are decorated with SV2⁺ contacts (arrow in bottom row), inset: unlesioned situation. Bars = 25 µm.

3.1.3.6 Evidence for motor axon growth out of the spinal cord

To determine whether newly generated motor neurons grow axons out of the spinal cord, we applied the retrograde neuronal tracer biocytin to the muscle tissue surrounding the lesion site of BrdU injected animals. Biocytin tracing

marks muscle-innervating neurons in the spinal cord, which are bona fide motor neurons. Fish were injected at 12, 13 and 14 dpl with BrdU and biocytin was applied at 42 dpl. Out of 4 fish, we found one BrdU⁺/biocytin⁺ cell (Fig.12), indicating that this newly generated cell extended an axon out of the spinal cord. The ventro-lateral position of the cell in the spinal cord is consistent with a motor neuron identity of this cell. One reason for the scarcity of these double-labelled cells may be that BrdU labels only a sub-population of newly generated motor neurons (approximately 25% at 14 dpl) and retrograde tracing does not label all motor neurons, such that overlap of the two markers may be a rare event. However, this observation suggests that newly generated motor neurons are capable of regenerating a peripheral axon.

Fig.12: Retrograde tracing of a newborn motor neuron in the spinal cord from the muscle periphery. Confocal images of the same spinal cross-section are shown in low (left) and high (right) magnification (dorsal is up). Arrows point to the same

biocytin/BrdU doubled labelled motor neuron at 8wpl. Bars = 50 µm (left), 15 µm (right).

3.1.4 Olig2:GFP⁺ ependymo-radial glial cells are potential motor neuron